The present study is naturalistic with a cross sectional design. It is organized as a substudy in the larger Thematically Organized Psychosis research (TOP) study. Data from the TOP study was used in two of the three studies (paper II and III) in this thesis. Data for the first study (paper I) was drawn from an existing database of a research program of the Veterans Affairs Connecticut Healthcare System and the Connecticut Mental Health Center, USA (1995-2002). The reader is referred to paper I for more detailed description of the American study.
The TOP study is an ongoing translational research study in Oslo, Norway aiming at investigating a range of issues associated to psychotic disorders from biological characteristics to clinical mechanisms. The TOP study is affiliated to the University of Oslo and University Hospitals in the Oslo area and participants are enrolled from mental health services including both in- and outpatients-units. The Norwegian health care has a system where patients are admitted by catchment area, i.e. all people are offered mental health care when needed within a given catchment area. This system allows for a high degree of patient representativity. The healthy controls used in paper II were randomly drawn from the population registers for the same catchments areas in Oslo as the patients and were contacted by letter with the request to participate. A screening process was conducted beforehand with an interview concerning severe mental illness, substance abuse and the Primary Care Evaluation of Mental Disorders (Spitzer et al., 1994). The TOP study has been approved by the Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate.
3.1.1. Procedure
Data from the American participants (paper I) was collected between 1995 and 1999 and between 1999 and 2002 as part of a vocational rehabilitation study program. Informed written consent was based upon procedures approved by the IRB at the VA Connecticut Healthcare System and participants completed The BORRTI and the PANSS as part of the intake measures.
Data for the TOP project was collected from clinical patients (n = 106) and healthy controls (n = 158) participating in the Thematically Organized Psychosis research (TOP) study in Oslo, Norway.
The clinical participants were referred to the TOP study on the suspicions of severe mental illness, mainly schizophrenia and bipolar disorders from their treatment units. Clinical and
neuropsychological data were collected along with structural and functional MRI and genetic information. Trained psychologists and psychiatrists carried out the clinical interviews under supervision of experienced psychiatrists specialized in diagnostics. Psychologists under supervision of specialized neuropsychologists conducted the neurocognitive assessments. Evaluation of symptoms (the PANSS) and information about object relations and reality testing (the BORRTI) was collected either at baseline or at six months follow-up. The PANSS and the BORRTI were
administered concurrently. If this was not possible, they were administered within maximum one week of one another.
Healthy controls from the same catchments areas as those of the treatment units were invited to participate in the TOP project by letter. The people who then responded received a phone call with questions regarding exclusion criteria. Assessments of object relations and reality testing (BORRTI) were carried out when the healthy controls were administered the neurocognitive assessments. For some participants this procedure was not carried out. They received the BORRTI questionnaire by letter and returned their replies by letter. Since this thesis was part of the Norwegian TOP study, I participated in the collection of data for paper II and III by carrying out
neurocognitive testing, clinical interviews and symptom evaluation of about a third of the patients in this study.
3.2. Participants
3.2.1. The American cohort
The American cohort consists of two hundred and seventy three outpatient participants from the mental health service of the VA Connecticut Healthcare System or the Connecticut Mental Health Center. They completed intake measures as part of a vocational rehabilitation study program.
Data was collected from 122 outpatients enrolled between 1995 and 1999 and from 151 outpatients enrolled between 1999 and 2002. The study is affiliated to The Yale University, School of Medicine and the VA Connecticut Healthcare System and Connecticut Mental Health Care Center, US. In addition to personality, symptoms, and insight measures, all participants were administered the BORRTI, (Bell, 1995) and the PANSS; (Key, Fizbein, & Opler, 1987) within the same period. All participants were diagnosed with schizophrenia or schizoaffective disorder and met the following criteria: no documented neurological disorder or development disability; GAF score over 30; no change in medication in the last 30 days. Eighty-seven percent (87 %) of the participants were male, 63 % were white, 32 % were African American and 4 % were Hispanic. Mean age at inclusion was 43.1 years, mean education was 13.0 years, mean age of illness onset was 22.6 years and mean duration of lifetime hospitalizations was 9.7 years.
3.2.2. The Norwegian cohort
The Norwegian cohort consists of 106 patients and 158 healthy controls (paper II). The patients (schizophrenia n = 55), (bipolar disorders n =51) were recruited to the study through their participation in the ongoing Thematically Organized Psychosis (TOP) Study, from in- and outpatient
units of the University Hospital of Oslo, Norway. The overall inclusion criteria for the TOP study were: age between 18 and 65 years, diagnosis within the psychosis spectrum disorders (DSM-IV);
schizophrenia, schizophreniform, schizoaffective disorder, psychosis not otherwise specified (NOS), delusional disorder, brief psychosis, major affective disorder with mood incongruent psychotic symptoms and bipolar disorder. Patients with neurological disorder, history of head injury, IQ<70, were excluded. In addition participants were required to comprehend Norwegian language at an acceptable level. Further inclusion criteria for the present study were, diagnosis within the schizophrenia spectrum disorders (schizophrenia and schizoaffective) and bipolar disorder (bipolar I and bipolar II).
Healthy controls (n = 158) were contacted by letter with the request to participate and were randomly drawn from the population registers for the same catchments areas in Oslo as the patients. A screening process was conducted beforehand with an interview about severe mental illness, substance abuse and the Primary Care Evaluation of Mental Disorders (Spitzer et al., 1994).
Trained psychologists and masters of neurosciences conducted the screening. The exclusion criteria’s were mental retardation (IQ<70), a history of head injury or difficulty speaking and understanding the Norwegian language. In addition, participants were excluded if they or any first-degree relative had a lifetime history of severe mental disorders, or if they had ongoing substance abuse in the last 6 months.
3.3. Measurements
3.3.1. Assessments of diagnosis
The diagnoses of the American cohort were based on the Structured Clinical Interview for DSM-III-R or DSM-IV (American Psychiatric Association, 1994b). Trained clinicians interviewed the participants. In the Norwegian cohort, diagnostic evaluations were based on the Structured Clinical
Interview for DSM-IV (First, Spitzer, Gibbon, & Williams, 1997). Trained psychiatrists and clinical psychologists, who had completed training course in SCID assessment and were regularly supervised on consensus meetings to assure high inter-rater reliability, carried out interviews. Diagnostic agreement was found satisfactory and the mean overall kappa based on the training program at UCLA (Ventura J, Libermann RB, Green MF, Shaner A, & Mintz J, 1998) was 0.77 (95% CI 0.60-0.94).
In paper II a separate variable for measuring the diagnosis of both schizophrenia and bipolar disorder was created in order to investigate their relationship to each of the BORRTI subscales.
3.3.2. Assessments of object relations and reality testing
The Bell Object Relations and Reality Testing Inventory (BORRTI) (Bell, 1995) is a self-report inventory consisting of 90 descriptive true or false statements answered according to the
respondent’s most recent experience. 45 items are measuring object relations and 45 measures reality testing – divided into the seven dimensions. Scoring yields the four factor-analytically derived object relations subscales – Alienation, Insecure Attachment, Egocentricity and Social Incompetence – and the three reality testing subscales – Reality Distortion, Uncertainty of Perception and Hallucinations and Delusions. Development of the inventory and definitions of the seven subscales are thoroughly described in the introduction above.
The inventory can be used to separately measure object relations part of the BORRTI (Form O) if investigating this function in non-psychotic persons (See Appendix). On the BORRTI, lower scores as well as higher scores can represent pathological features depending on the scale. For most scales higher scores indicates more pathology, but for Insecure Attachment and Uncertainty of Perception lower scores are considered pathological in persons with mental disorders. A low score on Insecure Attachment indicates insensitivity and indifference to relationships, - and a low score on Uncertainty of Perception in someone with schizophrenia is strongly associated with poor insight.
Psychometric studies of the BORRTI demonstrates good reliability and validity (Bell, 1995). A recent
study on validity of self-report in schizophrenia patients with poor insight and the BORRTI, showed limitations on the ability of these patients to accurately report on the accuracy of the perceptions, but otherwise there were external support for the validity of the BORRTI subscales (Bell, Fiszdon, Richardson, Lysaker PH, & Bryson G, 2007).
The BORRTI has shown good psychometric properties. The reliability of the instrument was evaluated by calculations of internal consistency, split-half reliability, test-retest reliability and stability of classifications. Internal consistency for each of the seven subscales (n = 336) was satisfactory (Cronbach’s Alpha was within the range of 0.79-0.90 and Spearman Split-Half between 0.77-0.90). The test-retest calculations (that evaluates the degree to which a respondent’s score remains stable over time) was assessed over 4, 13 and 26 weeks time and showed not too high or too low test-retest correlations for each scale. Because of the assessment was conducted in clinical groups undergoing treatment, and that the instruction to describe “your most recent experience”
may change the mental state of the respondent, fluctuation in the correlations may occur, which it did. Good reliability of classifications was found as the same deficits were found after re-evaluation at 6 months in a schizophrenia sample.
Validity of the clinical constructs that the BORRTI is measuring is much more difficult to determine than the instrument reliability. The instrument must be used a number of times and across a variety of respondents and settings to establish this. However, the construct validity of the BORRTI has included three components: theoretical-substantive, structural and external validity. The theoretical-substantive and structural validity was obtained in the development process and found satisfactory as far as it can be established. The external validity was evaluated by the convergence of experience derived from using the instrument in a variety of settings and studies. This indicates the BORRTI’s importance as a measure of individual differences and denotes the robustness of the underlying construct. (Bell, 1995). The ability of the BORRTI to discriminate among well-defined diagnostic groups was evaluated by comparing the group-mean of their BORRTI scores in several
studies, - and was found satisfactory. Also the relationship to other personality and clinical measurements (Brief Psychiatric Rating Scale (BRPS) (Overall & Gorham, 1962), The Global Assessment Scale (GAS) (Endicott, Spitzer, Fleiss, & Cohen, 1976) and the Positive And Negative Syndrome Scale (PANSS) (Kay et al., 1987) showed good convergent and divergent validity (Bell, 2004). This was confirmed in a review of the BORRTI that concluded that it is a reliable and valid assessment of object relations and reality testing (Alpher, 1990).
In order to investigate object relations and reality testing functioning in the Norwegian cohort of this study, the BORRTI was translated into Norwegian. The translation was validated using the back-translation method (Brislin, 1970) and the Norwegian version of the BORRTI was translated back to English by a bilingual psychologist, and approved by the author of the original inventory. The Norwegian translation of the BORRTI showed good psychometric properties. The Cronbach’s alpha seems satisfactory for both the Norwegian and US BORRTI subscales and when compared they had high inter-correlations Furthermore, the Norwegian translation of the BORRTI showed good discriminant validity, as significant group differences were found between the two diagnostic groups and healthy controls included in the study on almost all the subscales (Hansen, Torgalsbøen, Bell, &
Melle, 2012).
3.3.3. Assessments of Passive Social Withdrawal and Active Social Avoidance
In order to assess observed social withdrawal we used The Positive and Negative Symptom Scale (PANSS); (Key et al., 1987), a 30 item rating scale comprising a wide range of positive, negative and general psychopathological symptoms. It is scored after a semi-structured interview and rated from 1 (not present) to 7 (extremely severe) using behavioral anchors based upon the last seven days. In paper I and III we isolated two items on the Positive And Negative Syndrome Scale (PANSS), (Kay et
al., 1987); N4 (Passive/Apathetic Social Withdrawal) on the negative symptom scale, and G16 (Active Social Avoidance) on the general psychopathology scale.
The differences between the two items have been found in most factor analyses of the PANSS. Without exception the N4 loads on the negative components, the G16 Active Social Avoidance has found to load on several factors, a depression-anxiety factor (Bell, Lysaker, Beam-Goulet, & Milstein, 1994), negative, excitement, emotional distress and positive factor (van der Gaag et al., 2006) and not on any factor at all (White, Harvey, Opler, & Lindenmayer, 1997).
The validity of PANSS scores on the item level was recently evaluated by Santor et al. (2007) using item response theory (IRT). This explores the performance of each item on the PANSS regarding their effectiveness to discriminate among individual differences in symptom severity and the appropriateness of cutoff scores. Each item went through an analysis of response within an acceptable region on option characteristic curves. The regions are created based on an overall total score of the sample and both the N4 Passive Social Withdrawal and G16 Active Social Avoidance were found “very good” (Santor, Ascher-Svanum, & Obenchain, 2007).
The PANSS was rated by clinically trained research staff and the American inter-rater reliability for raters were in the excellent range for the five component scores (ICC = 0.88 to 0.93) of the five factor model that was used (Bell et al., 1994). The Norwegian cohort was also rated by clinically trained research staff and the inter-rater reliability was good: intra-class correlation coefficient (ICC 1.1), for the Positive subscale: 0.82 (95% CI0.66-0.94), the Negative subscale: 0.76 (95% CI 0.58-0.93), and General subscale 0.73 (95% CI 0.54-0.90). In paper II we used the Five Factor-Model of schizophrenia based on the factorial invariance of the PANSS (Bell et al., 1994), in order to be able to compare with previous BORRTI studies. The components are: Positive component (unusual thought content, delusions, suspiciousness, grandiosity, hallucinatory behavior, somatic concern); Negative component (passive withdrawal, emotional withdrawal, blunted affect, preoccupation, lack of spontaneity & flow of conversation, poor rapport, motor retardation,
disturbance of volition); Cognitive component (conceptual disorganization, poor attention, tension, difficulty in abstract thinking, lack of judgment & insight, stereotyped thinking); Emotional Discomfort component (anxiety, guilt, depression and active social withdrawal) and Hostility component (hostility, poor impulse control, uncooperativeness, excitement). The Emotional Discomfort component is used to measure current depressive symptoms in paper II.
3.3.4. Assessments of subjective experience of social withdrawal
Subjective experienced social withdrawal was evaluated using the Social Functioning Scale (SFS), (Birchwood et al., 1990). Two subscales were isolated: SFS Withdrawal (time spent alone, social avoidance and conversation initiative); and the SFS Interpersonal Behavior (romantic involvement, number of friends and quality of communication). The SFS is a self-report questionnaire measuring; Withdrawal, Interpersonal Behavior, Prosocial Activities, Recreation, Independent Competence, Independent Performances and Employment. Scaled Scores (normalized) have a mean of 100 and standard deviation of 15. The Norwegian translation has shown good reliability and validity (Hellvin et al., 2010).
3.3.5. Other measurements
In the TOP project, information about history of mental illness, present symptoms and pharmacological treatment were collected by interview with the patients. Information was also gathered from treatment records and clinical staff. To evaluate current level of symptoms The Positive And Negative Syndrome Scale (PANSS) (Kay et al., 1987), the Young Mania Rating Scale (YMRS) (Young, Biggs, Ziegler, & Meyer, 1978) and the Inventory of Depressive Symptoms (Clinician rated) (IDS ದ C), (Rush, Gullion, Jarrett, & Trivedi, 1996) were included in the general TOP protocol.
For the three studies in this thesis, we used the Five Factor-Model of the PANSS of schizophrenia
(Bell MD., Lysaker, Beam-Goulet, & Milstein, 1993) to be able to compare with previous BORRTI studies. The PANSS Positive component (unusual thought content, delusions, suspiciousness, grandiosity, hallucinatory behavior, somatic concern) was used, for evaluating current level of positive symptoms. For assessment of the current level of depression, we used the Emotional Discomfort component (anxiety, guilt, depression and active social withdrawal) and Hostility component (hostility, poor impulse control, uncooperativeness, excitement).
In paper II, several other variables were included. In order to explore the role of lifetime history of psychosis and its relationship to BORRTI subscale scores, we created a variable measuring whether or not each participant had had a lifetime psychotic episode based on information from the SCID interview. No episode of psychosis during lifetime includes only patients with bipolar disorder, since the schizophrenia diagnosis requires the presence of psychotic features.
In paper III we needed to ensure that combining the two diagnostic groups in one analysis was valid. We therefore created several variables with the BORRTI subscales and diagnosis. Then we conducted a series of linear regression analyses with the PANSS PSW, the PANSS ASA, the SFS Withdrawal and the SFS Interpersonal Behavior as dependent variables. The variables comprised each of the BORRTI subscales, the diagnostic groups, - and the interaction term between the BORTTI subscales and the diagnostic group. These were then used as independent variables. After ruling out any interaction effects for diagnosis, we continued with the main analyses in the combined patient sample.
3.4. Statistical analysis
All analyses were carried out using the Statistical Package for the Social Sciences version 16 for paper I and PASW version 18 for paper II and III (SPSS Inc., Chicago, IL, USA). Primary analyses were performed to ensure data quality for all variables including inspection of skewness, linearity and outliers. Descriptive statistics for both the American and Norwegian samples were obtained using standard deviations, means, medians or range according to the type of assessment.
Relationships between continuous variables were analyzed with Pearson’s correlation and the level of significance was set to p=0.05, two-sided. A range of multiple regression analysis was used to predict scores on a continuous variable (paper I), and analyses were done in a forward stepwise procedure with an entry criterion of p = 0.15. Age and gender were entered as covariates.
In paper II and III we merged the schizophrenia and bipolar disorder patients into one sample. In order to ensure that combining the two diagnostic groups in one analysis were valid (paper III), we did a series of linear regression analyses with the continuous assessments and the interaction term between diagnostic group and the BORTTI subscales as independents to rule out any possible interaction effects for diagnosis.
Raw scores of the BORRTI subscale were transformed into z-scores based on the norms of the Norwegian healthy control sample. To compare the BORRTI profiles across diagnostic groups we used a one-way ANOVA with Scheffe’s Post- Hoc corrections. For a more detailed description of the statistical analyses used in the three studies, the reader is referred to the method section of each of the papers.